Contents lists available at ScienceDirect

Molecular Phylogenetics and Evolution

journal homepage: www.elsevier.com/locate/ympev

Diversification of shrub frogs (Rhacophoridae, Pseudophilautus) in Sri Lanka– Timing and geographic contextMadhava Meegaskumburaa,b,1,⁎, Gayani Senevirathnec,1, Kelum Manamendra-Arachchid,Rohan Pethiyagodae, James Hankenf, Christopher J. Schneiderg,⁎

a College of Forestry, Guangxi Key Lab for Forest Ecology and Conservation, Guangxi University, Nanning 530004, PR ChinabDepartment of Molecular Biology & Biotechnology, Faculty of Science, University of Peradeniya, Peradeniya, Sri Lankac Department of Organismal Biology & Anatomy, University of Chicago, Chicago, IL, USAd Postgraduate Institute of Archaeology, Colombo 07, Sri Lankae Ichthyology Section, Australian Museum, Sydney, NSW 2010, AustraliafMuseum of Comparative Zoology, Harvard University, Cambridge, MA 02138, USAg Department of Biology, Boston University, Boston, MA 02215, USA

A R T I C L E I N F O

Keywords:Ancestral-area reconstructionBiogeographyEcological opportunityDiversificationMolecular datingSpeciation

A B S T R A C T

Pseudophilautus comprises an endemic diversification predominantly associated with the wet tropical regions of SriLanka that provides an opportunity to examine the effects of geography and historical climate change on diversi-fication. Using a time-calibrated multi-gene phylogeny, we analyze the tempo of diversification in the context ofpast climate and geography to identify historical drivers of current patterns of diversity and distribution. Moleculardating suggests that the diversification was seeded by migration across a land-bridge connection from India during aperiod of climatic cooling and drying, the Oi-1 glacial maximum around the Eocene-Oligocene boundary. Lineage-through-time plots suggest a gradual and constant rate of diversification, beginning in the Oligocene and extendingthrough the late Miocene and early Pliocene with a slight burst in the Pleistocene. There is no indication of an early-burst phase of diversification characteristic of many adaptive radiations, nor were there bursts of diversificationassociated with favorable climate shifts such as the intensification of monsoons. However, a late Miocene (8.8 MYA)back-migration to India occurred following the establishment of the monsoon. The back migration did not trigger adiversification in India similar to that manifest in Sri Lanka, likely due to occupation of available habitat, andconsequent lack of ecological opportunity, by the earlier radiation of a sister lineage of frogs (Raorchestes) withsimilar ecology. Phylogenetic area reconstructions show a pattern of sister species distributed across adjacentmountain ranges or from different parts of large montane regions, highlighting the importance of isolation andallopatric speciation. Hence, local species communities are composed of species from disparate clades that, in mostcases, have been assembled through migration rather than in situ speciation. Lowland lineages are derived frommontane lineages. Thus, the hills of Sri Lanka acted as species pumps as well as refuges throughout the 31 millionyears of evolution, highlighting the importance of tropical montane regions for both the generation and main-tenance of biodiversity.

1. Introduction

Mountainous islands in the tropics are often hotspots of endemismand diversity but the processes underlying this pattern are poorlyknown. The isolation of islands contributes to their high endemism, butthe processes of diversification within islands are complex. Interaction

between broad-scale climate fluctuations and the topographic com-plexity of mountainous tropical regions may promote the accumulationand persistence of species richness in a variety of ways (Fjeldsa et al.,2012; Graham et al., 2004; Smith et al., 2014). Habitat variation fromlow to high altitudes provides the opportunity for speciation due todivergent selection across ecological gradients (Cadena et al., 2011;

https://doi.org/10.1016/j.ympev.2018.11.004Received 8 July 2018; Received in revised form 7 November 2018; Accepted 10 November 2018

⁎ Corresponding authors at: College of Forestry, Guangxi Key Lab for Forest Ecology and Conservation, Guangxi University, Nanning 530004, PR China (M.Meegaskumbura) and Department of Biology, Boston University, Boston, MA 02215, USA (C.J. Schneider).

E-mail addresses: madhava_m@mac.com (M. Meegaskumbura), gsenevirathne@uchicago.edu (G. Senevirathne), kelum@gmail.com (K. Manamendra-Arachchi),rohanpet@gmail.com (R. Pethiyagoda), hanken@oeb.harvard.edu (J. Hanken), cschneid@bu.edu (C.J. Schneider).

1 These authors contributed equally to this work.

Molecular Phylogenetics and Evolution 132 (2019) 14–24

Available online 16 November 20181055-7903/ © 2018 Elsevier Inc. All rights reserved.

T

Graham et al., 2004; Ogden and Thorpe, 2002; Smith et al., 2001). Thehigh relief of montane regions, with cool, wet upland zones often se-parated by warmer and drier valleys or lowlands, provides the oppor-tunity for allopatric divergence of cool-wet-adapted species on isolatedmountain ranges (Bossuyt et al., 2004; Cadena et al., 2011; Kozak andWiens, 2006; Meegaskumbura and Manamendra-Arachchi, 2005;Moreau, 1966; Schneider et al., 1998; Schneider and Moritz, 1999;Vijayakumar et al., 2016). In addition, montane regions can provideclimatically stable areas where suitable habitats persist through globaland regional climate fluctuations (Carnaval et al., 2009, 2014; Fjeldsaet al., 2012; Schneider and Williams, 2005), thus contributing to thebuildup of species richness. Teasing apart the relative contributions ofclimate, geography and different evolutionary processes on diversifi-cation is a longstanding challenge. Recent advances in phylogenetic andhistorical biogeographic analyses, however, provide the opportunity togain insight into the pattern and process of diversification that allowassessment of the relative effects of geology, geography, dispersal andclimate on diversification, persistence and species richness.

Rhacophorid frogs of the genus Pseudophilautus in Sri Lanka re-present a spectacular, endemic diversification of anurans on a moun-tainous continental island (Manamendra-Arachchi and Pethiyagoda,2005; Meegaskumbura et al., 2002; Meegaskumbura and Manamendra-Arachchi, 2005; Meegaskumbura, 2007). Seventy-six extant species arerecognized at present, based on both morphological and molecular data(AmphibiaWeb, 2018), representing a five-fold increase from the ap-proximately 15 species recognized less than 20 years ago. The re-markable diversification of these frogs appears to have been aided bytheir direct-developing reproductive mode, which allows them to breedin moist habitats away from bodies of standing or running water, evenin relatively dry areas (Bahir et al., 2005). Similar diversifications haveoccurred in other direct-developing anuran genera such as the Ter-rarana clade of the New World (Hedges et al., 2008) and the Indianbush frogs, Roarchestes (Vijayakumar et al., 2016), and direct devel-opment appears to be a key innovation that facilitates diversification(Blackburn et al., 2013; Bossuyt and Milinkovitch, 2001; Setiadi et al.,2011). In Sri Lanka, Pseudophilautus species occupy several habitattypes ranging from open grasslands, rock substrates along streams,rainforest canopy, shrubs, leaf litter and anthropogenic habitats, andmost are confined to higher elevations of the wet, montane region of theisland’s southwestern quarter (Manamendra-Arachchi and Pethiyagoda,2005; Meegaskumbura et al., 2007, 2009, 2012; Meegaskumbura andManamendra-Arachchi, 2005, 2011). The three main mountain rangesof this region constitute ecological islands of cool, wet habitat separatedby low, warm and often drier valleys.

Key questions regarding diversification of Pseudophilautus in SriLanka following the initial colonization concern the geographic contextof speciation and the impact of climate change over time. Are speciesassemblages in each mountain range monophyletic or is there evidencefor dispersal between mountain ranges? Did diversification occur as anearly burst or as a succession of events of geographic range expansionand subsequent fragmentation associated with climatic cycles? Arethere geologic and/or climatic events that had overarching impacts andshaped the radiation over time?

Here we use a multi-gene, time-calibrated phylogeny of nearly allextant species of Sri Lankan and Indian Pseudophilautus to estimate ratesof diversification over time in relation to key geologic and geographicevents and to evaluate hypotheses regarding the drivers of patterns ofdiversity.

2. Materials and methods

2.1. Phylogeny

2.1.1. Tissue collection, extraction, PCR amplification and sequencingThis analysis included 67 putative species of Pseudophilautus in Sri

Lanka; 2 species of Raorchestes, the sister group of Pseudophilautus

(Abraham et al., 2013)—R. charius and R. signatus; the Indian ingroupspecies Pseudophilautus wynaadensis, P. amboli and P. kani; and 1 or 2representatives from each of the other rhacophorid genera (S1 Table).The trees were rooted using 4 species of Mantellidae: Mantella aur-antiaca, M. madagascariensis, Mantella sp. and Mantidactylus grandidieri.Previous studies suggest that this family is the closest sister-group ofRhacophoridae (Li et al., 2013; Meegaskumbura et al., 2015). The 67putative species of Pseudophilautus mentioned above were included toderive as complete a phylogeny as possible for the Sri Lankan diversi-fication. Of these species, 48 have been formally described and named,whereas 19 await formal description. Tissue samples were collected inthe field, along with locality data. (See S1 Table for taxon sampling,gene sampling and distributional data, as well as reference numbers.)

For molecular systematic analysis, DNA was extracted from ethanol-preserved tissues using Qiagen tissue extraction kits following manu-facturer’s protocols. A total of 2907 base pairs (bp) were sequenced,which included two mitochondrial ribosomal RNA gene fragments (12SrRNA and 16S rRNA; ca. 1504 bp) and one large fragment of the nucleargene Rag-1 (ca. 1403 bp). Portions of the 12S rRNA, 16S rRNA and Rag-1 genes were amplified by PCR and sequenced directly using dye-ter-mination cycle sequencing. Primer sets and protocols for both PCR andsequencing are described in detail in the Supplementary Material (S1Text).

2.1.2. Sequence alignmentChromatograms were edited visually using 4peaks (v. 1.7.1).

Mitochondrial 16S rRNA and 12S rRNA gene fragments were initiallyaligned using ClustalW as implemented by MEGA v.6.0 (Tamura et al.,2013); regions for which we had low confidence in positional homologywere removed from the analysis. Nuclear Rag-1 gene sequences werealso aligned using MEGA v.6.0 with translated amino-acid sequences.The complete concatenated dataset included 94 taxa (duplicated taxawere pruned from the dataset) with a total of 2177 bp (16S rRNA,464 bp; 12S rRNA, 310 bp; Rag-1, 1403 bp).

2.1.3. Phylogenetic analysesTree topology was inferred using maximum likelihood (ML) ana-

lyses that were performed for both partitioned and unpartitioned da-tasets and also for each gene individually. The dataset was partitionedinto specific gene regions by specifying character sets (charset 16SrRNA=1–464; charset 12S rRNA=465–775; charset Rag-1=776–2178). The partitioned dataset was used for the phylogeneticanalyses discussed in detail. The best-fitting nucleotide substitutionmodel for each dataset was chosen using jModelTest v.2.1.4 (Darribaet al., 2012; Guindon and Gascuel, 2003). We used the Cipres ScienceGateway Server (Miller et al., 2010) to run the ML analysis usingRAxML 7.2.8 (Stamatakis, 2006)) on the partitioned dataset. The modelof evolution, as selected from the jModelTest (GTR+ I+G), was givenwith all parameters estimated along with 1000 bootstrap pseudor-eplicates (model parameters for each dataset are given underSupplementary Information). Subsequently, the unpartitioned datasetand the three genes individually were also analyzed using RAxML.

Clade support was assessed using ML bootstrapping and Bayesianposterior probabilities (PP). Bootstrapping was done in an ML frame-work using RAxML 7.2.8 (Stamatakis, 2006), with 1000 RAxML sear-ches and 1000 iterations. BEAST v.1.8 (Drummond and Rambaut,2007) was used to infer trees with a Bayesian criterion and to assessposterior probability at each node. The Yule model was given as the treeprior; the model of sequence evolution was the same as that used in theML analysis. Four Metropolis-Coupled Markov Chain Monte Carlo(MCMCMC) chains were run for 10 million generations and for twoconsecutive runs. Burnin was defined by observing the log-output file inTracer v.1.6 (Drummond et al., 2012); 90% of the post-burnin treeswere analyzed using TreeAnnotator. Consensus trees using unparti-tioned datasets for ML and Bayesian analyses are given asSupplementary Information, as are log-output files.

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

15

Analyses of partitioned and unpartitioned datasets yielded similartopologies but slightly different branch lengths. However, since thepartitioned data set provides greater node support, we use this datasetand the corresponding tree for timing-related analyses and geographicreconstructions.

2.2. Divergence time

To examine the temporal context of divergence and to correlate itwith geological and climatic events, we estimated divergence timeamong lineages on the partitioned dataset by using BEAST v.1.8(Drummond and Rambaut, 2007). The Yule model was given as the treeprior, the GTR+ I+G model of evolution specified, and other para-meters estimated for all gene partitions. Over 10 million generations ofMCMC simulations were run, with one tree saved per 1000 generations.The analysis was repeated twice: once using the lognormal relaxedclock (uncorrelated) and a second time using the default strict clock(with random starting trees). The lognormal relaxed clock modelshowed the greatest fit and was used in all subsequent analyses. Wecalibrated the tree using the split between Mantellidae and Rhaco-phoridae, a reference point used in previous studies (Li et al., 2013;Roelants et al., 2004), using the fossil calibration date (73.1 ± 19MYA) designated by Bossuyt and Milinkovitch (2001).

2.3. Lineage-through-time (LTT) analysis

We analyzed the accumulation of lineages through time to discernthe tempo of diversification in the radiation (Harmon et al., 2007) andto assess its possible relation to climatic and geographic events.Lineage-through-time plots were constructed by plotting the lognumber of lineages against the log divergence time. The plots weregenerated in R v.1.0 using the packages ape, laser and geiger (Paradis,2006; Rabosky, 2006). A null model was also developed wherein thetree was simulated with a total of 94 taxa and assuming a constant rateof speciation across the lineages (under a pure birth model) with noextinctions. A Monte Carlo Constant Rate test (MCCR; Pybus andHarvey, 2000) was used to calculate the gamma (γ) statistic (using thelaser package; Rabosky, 2006), which compares the distances betweenthe tree-root and the midpoint to the distances between the midpointand tips. If γ < 0, the radiation shows an early, rapid speciation event;if γ= 0, the radiation undergoes a constant diversification rate; and ifγ > 0, the diversification rate increases towards the present (Pybusand Harvey, 2000).

2.4. Geographic reconstruction

To evaluate the geographic origins and movements of lineages overtime, we estimated the ancestral geographic state at nodes in the phy-logeny using both ML (Lagrange v. 1.5; Ree and Smith, 2008) andBayesian frameworks (BEAST v.2.0; Bouckaert et al., 2014).

The Lagrange analysis included the following steps: the partitioneddataset and the chronogram resulting from the BEAST dating analysiswere uploaded to the Lagrange Configurator. Sri Lankan species in-habiting elevations 700m and above were segregated by the three mainmountain ranges: Central Hills (C), Knuckles Hills (K) and RakwanaHills (R). Species found below 700m were placed in a single category:Lowlands wet (LL). Indian species were coded as (IND) and the re-maining rhacophorids+ outgroup were coded as undesignated (U), astheir distributional data were not included in the analysis. Based onavailable data on the distribution of Sri Lankan species (AmphibiaWeb2018; Bahir et al., 2005; Meegaskumbura and Manamendra-Arachchi,2005, 2011; Manamendra-Arachchi and Pethiyagoda, 2006;Meegaskumbura et al., 2007, 2009, 2012; Wickramasinghe et al.,2013), the following dispersal routes were evaluated: Central-Rakwana,Central-Knuckles, Central-Lowlands wet, Central-India, Knuckles-

Rakwana, Knuckles-Lowlands wet, Knuckles-India, Rakwana-Lowlandswet, Rakwana-India and India-Lowlands wet. Assuming that Pseudo-philautus have poor dispersal abilities over non-adjacent areas, lowprobability values (0.1) were assigned to four routes: Central-India,Rakwana-India, Knuckles-India and Knuckles-Rakwana. The maximumnumber of areas reconstructed at a node was fixed as two.

Reconstructions using the Bayesian framework were done using theStandard-BEAST method as implemented in BEAST v. 2.2.0 (Bouckaertet al., 2014). Codings for the different regions (C, R, K and U) weredefined in the same way as in the Lagrange analysis. The input XML filewas created using BEAUti v. 2.0 using the partitioned dataset. Thelognormal-relaxed (uncorrelated) clock was adopted, a Yule speciationmodel was given as the tree prior, and a gamma prior distribution wasused for the discrete trait. The run time was ten million generations andthe MCMCMC analysis was performed on four chains. The burnin pointwas determined by observing the log file on Tracer v1.6 (Rambautet al., 2014) and 90% of post-burnin trees were summarized inTreeAnnotator.

3. Results

3.1. Phylogeny

Phylogenetic analysis strongly supports the monophyly ofPseudophilautus, with the clade of Raorchestes+Mercurana as its sistergroup. Internal nodes of the Pseudophilautus clade are for the most partwell supported. Building on these relationships, we identify six compo-nent clades (Fig. 1, clades A–F). Clade A mostly comprises species thatinhabit higher elevations of the three major mountain ranges. Clade B hasa broad representation of species from low to high elevations. Clade C ismostly an assemblage of species from low to high elevations and includessome of the largest species of Pseudophilautus. Clade D harbors frogs fromlow to high elevations. Clade E is represented mostly by low-to-mid-ele-vation species, but it also includes a clade harboring three Indian species.Clade F comprises some of the smallest frogs, which inhabit the higherelevations of the three major mountain ranges. The genera Buergeria,Liuixalus, (Theloderma+Nyctixalus), ((Rhacophorus+Ghatixalus+ Poly-pedates+ Taruga)+ (Feihyla+“Feihyla” + Chiromantis)), (Gracixalus),(Philautus), (Kurixalus) and (Beddomixalus), (Mercurana+Roarchestes),and (Pseudophilautus) formed sequential sister clades (Fig. 1). Mercuranaor Raorchestes is the sister group of Pseudophilautus; the node uniting(Raorchestes+Mercurana) is poorly supported by ML bootstrap (64) buthighly supported by Bayesian posterior probability ( > 95). Topologiesof the consensus trees of Bayesian and ML analyses are similar in theirassembly of major clades (Figs. 1, 2).

3.2. Divergence-time and Lineage-through-time analyses

Divergence-time analysis, calibrated using the rhacophorid-man-tellid split (dated at 67.8 MYA; Li et al., 2013), suggests that the splitbetween Pseudophilautus and Raorchestes occurred ca. 31 MYA and thatthe most recent common ancestor (MRCA) of extant Pseudophilautusdates to ca. 25 MYA (95% PI= 21.93–45.14; Fig. 2; Table S4). Majorclade diversification seems to have occurred during the early-to-midMiocene, from 21.3 to 13.4 MYA (Fig. 2A). However, most extantPseudophilautus species evolved during the late Miocene and Pliocene,and several currently recognized species evolved during the earlyPleistocene (ca. 1 MYA). The back-migration of Pseudophilautus to Indiaoccurred during the late Miocene (Fig. 2C).

Lineage-through-time plots suggest that diversification withinPseudophilautus occurred gradually throughout the last 31 million years(Fig. 2B); there is no indication of an early burst of speciation duringthe initial colonization of the island. The log-null model (a pure birthmodel with constant diversification rates across the lineages) is close tothe actual cumulative lineage curve (Fig. 3). The γ statistic value of

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

16

1000 MCCR simulations, 2.91, suggests there was a low magnitude ofdiversification after the midpoint of the tree. Based on the slight de-viation from the null model (Fig. 3), the slightly positive value can beattributed to the lineage’s late diversification (late Pliocene and Pleis-tocene).

3.3. Geographic area reconstruction

Ancestral-area reconstructions using Lagrange (ML approach) andBEAST (Bayesian approach) yield similar patterns. For brevity, we de-pict the Lagrange analysis (Fig. 2); the BEAST reconstruction is

Fig. 1. Multi-gene (12S rRNA, 16S rRNA and Rag-1) partitioned ML phylogram of rhacophorid genera with an emphasis on Sri Lankan-Indian species ofPseudophilautus. The tree includes a total of 94 taxa and is rooted with four mantellid species. The monophyletic radiation of Pseudophilautus comprises six clades,A–F. Its sister group is a clade containing the Indian genera Mercurana and Raorchestes. This topology forms the basis of all subsequent analyses. Asterisks (*) aboveand below nodes indicate> 95% Bayesian posterior probabilities and ML bootstrap values, respectively.

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

17

23.7

67.8

13.4

25.3

19.3

31.1

22.5

14.8

21.3

16.9

2 8

63.5

21.0

Ps. stellatus HFS01002Ps. frankenbergi WHT2552Ps. cf. frankenbergi WHT2729Ps. femoralis WHT2772Ps. cf. mooreorum WHT6306Ps. poppiae WHT2779Ps. cf. poppiae WHT5051Ps. mooreorum WHT3209Ps. sarasinorum WHT2481Ps. procax WHT2786Ps. abundus WHT3231Ps. cf. schmarda WHT2501Ps. hankeni WHT6302Ps. schmarda WHT2715Ps. lunatus WHT3283Ps. viridis WHT2766Ps. stuarti WHT3208Ps. hallidayi WHT2886Ps. hallidayi WHT2805Ps. cavirostris WHT3299 Ps. cf. cavirostris WHT6381Ps. popularis WHT3191Ps. cf. popularis WHT3224Ps. cf. popularis WHT6074Ps. alto WHT5029Ps. cf. popularis WHT6010Ps. cf. macropus WHT2484Ps. sordidus WHT2699Ps. cf. sordidus WHT2795Ps. cf. sordidus WHT2796Ps. steineri WHT3210Ps. microtympanum WHT2558Ps. cf. microtympanum 6305Ps. cf. silus1 WHT2489Ps. cf. silus3 WHT6070Ps. fulvus WHT3121Ps. cf. silus2 WHT3188Ps. silus WHT6313Ps. reticulatus WHT3230Ps. papillosus WHT3284Ps. caeruleus WHT2511Ps. hoipolloi WHT2675Ps. pleurotaenia WHT3176Ps. hoffmanni WHT3223Ps. asankai WHT5107Ps. ocularis WHT2792Ps. auratus WHT2887Ps. decoris WHT3271Ps. mittermeieri WHTKAN2Ps. singu WHT6034Ps. tanu WHT6343Ps. silvaticus WHT2515 Ps. cf. limbus WHT2690Ps. limbus WHT2700Ps. cf. limbus WHT2540Ps. rus WHT5871Ps. schneideri WHT2667Ps. stictomerus WHT3301Ps. zorro WHT3175Ps. cuspis WHT5974Ps. cf. folicola1 WHT2525Ps. folicola WHT6114Ps. cf. folicola2 WHT2531Ps. cf. folicola3 WHT2915Ps. regius WHT2669Ps. wynaadensisPs. amboliPs. kaniPs. cf. simba WHT3221Ps. simba WHT6004Ps. semiruber WHT5831Raorchestes chariusRa. signatusMercurana myristicapalustris Beddomixalus bijuiKurixalus eiffingeriKuri odontotarsusPhilautus aurifasciatusGracixalus gracilipesChiromantis doriae Chiromantis vittatus Feihyla palpebralisPolypedates leucomystaxTaruga eques WHT2741Ghatixalus variabilisRhacophorus reinwardtiiNyctixalus pictus Theloderma rhododiscusLiuixalus romeriBuergeria oxycephalus

Mantella aurantiacaMantella sp.

Mantydactylus madagascariensisMantidactylus grandidieri

186

185

183

182

181

180

178

176

172

170

164

163

162

126

8.8

123

119120

124

115116

150152

156157

159

160

145

138

141142

146

147133

131134135

105104

103106

110

109

8793

97

96

9290

111

CretaceousLate

PaleogenePaleocene Eocene Oligocene

NeogeneMiocene Epoch

PeriodQuartenary

Pleisto

cene

Late Paleocene Thermal Maximum

Oi-1 Glaciaiton Mi-1 GlaciaitonAsian Monsoons Intensify (Himalaya Formidable)

Antarctic Ice Sheets Northern Hemisphere Ice Sheets

Himalayan-Tibetan orogeny - initiationIndian plate collideswith Eurasian plate

RCK700m

0

1000

2000

YXLL LL

50 100 150 200 0

LL

LL

CLL

R

K

X

Y50km

2000 mm

0 - 700 m> 700 m

Pseudophilautusback-migration to India8.8 MYA (LATE MIOCENE)

Pseudophilautus ancestormigrates from India 31.1 MYA (OLIGOCENE)

S R I L A N K A

Pseudophilautus diversificationon island with mountains acting as species pumps and refuges

1

I N D I A

2

1

2

AB

C

C- Central HillsK - Knuckles HillsR - Rakwana HillsLL - LowlandIND - IndiaUndesignated

CladeA

CladeB

CladeC

CladeD

CladeE

40

3741

42

34

2830

31

43

44

84

7071

7476

77

72

81

82

83

55

5256

64

63

62

60

4849

168169

175

1415

2021

22

23

10

24

6

MYA01 02 03 04 05 06 07 0

Full Scale & PermanentEphemeral

Plioce

neAltitude (m)

Distance along transect (km)

CladeF

161

Mi-4 Glaciaiton

EO

T

Pseudophilautusdiversification

08

Fig. 2. Timing and context of Pseudophilautus diversification. (A) Time-calibrated tree of the dating analysis using BEAST and output from the Lagrange analysis forarea reconstruction. Node numbers from the Lagrange analysis are set in black font (see also S4 Table). Numbers set in brown font at several major nodes are datingestimates in MYA (error intervals in Table S3). Key events include the common ancestor of Pseudophilautus migrating to Sri Lanka over exposed land bridges duringthe late Oligocene (perhaps at the Oi-1 glacial maxima); a late-Oligocene-to-Pleistocene diversification on the island, which may have been facilitated by theHimalayan-Tibetan orogeny and correlated intensification of the monsoons; and a late Miocene back-migration of Pseudophilautus to the mainland, likely across landbridges during a glacial period (perhaps at the Mi-4 glacial maxima). (B) Depiction of Pseudophilautusmovements between India and Sri Lanka. (C) A map of Sri Lankadepicting the three main mountain ranges (C, Central Hills; K, Knuckles Hills; R, Rakwana Hills), Lowlands (LL) and India (IND). The 2000mm rainfall contourdemarcates the wet zone in Sri Lanka. X–Y line depicts the altitudinal transect across the three main mountains.

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

18

presented in Supplementary Information (S1 Fig.). Reconstruction ofancestral geographic area places the MRCA of the Pseudophilautus ra-diation in the mountainous Central Hills (Fig. 2A; node 164). Sub-sequent colonization of areas outside the Central Hills has occurredrepeatedly in each major clade. Ancestral-area reconstruction alsoshows a frequent pattern of sister species distributed across the threemountain regions, as well as diversification within a given geographicregion (Fig. 2).

Clade A (Fig. 2A; node 126) is dominated by species from theCentral Hills, with the ancestor predicted to have been a Central Hillsform. The lineages in this clade are confined to cool, wet habitats athigher elevations (above 800m) and therefore are not likely to disperseamong mountain ranges. However, there is evidence for one dispersalevent from the Central Hills to the Rakwana Hills, and another to theKnuckles Hills, during the mid-late Miocene, resulting in the evolutionof distinct species. The species that resulted from dispersal to theRakwana and Knuckles ranges are most closely related to P. femoralis inthe Central Hills. Of all the Central Hills taxa in Clade A, P. femoralis hasthe lowest elevational range (currently occurring down to 800m abovesea level), suggesting that the P. femoralis lineage may have a greaterprobability of dispersal to other mountain regions than lineages re-stricted to the cooler, higher altitudes of the Central Hills.

Clade B (Fig. 2A; node 161) shows a more complex geographicpattern. Whereas the ancestor of the clade is estimated to have been inthe Central Hills, there are at least seven instances of dispersal fromthere to other mountain ranges, with P. hallidayi, P. sarasinorum and P.cavirostris widely distributed across both the lowlands and mountainranges up to an elevation of 1000m. The relationship between lineage-specific ecological characteristics and dispersal in this clade, and all

others, is complex and is elaborated more fully in the Discussion below.Clade C (Fig. 2A; node 111) is estimated to have arisen from a

Central Hills ancestor, but with early dispersal to the Rakwana andKnuckles Hills and one species (P. sordidus) that is widespread acrossboth lowland and mountain regions. Although the reconstructions forthis clade are ambiguous, it appears that there was minor diversifica-tion within both the Knuckles and Rakwana Hills, with a complexhistory of migration to and from the Central Hills.

Clade D (Fig. 2A; node 44) is reconstructed as arising from a CentralHills ancestor but with migration events to the Rakwana Hills and thelowlands followed by subsequent diversifications within those ranges([P. decoris, P. mittermeieri, P. tanu, P. singu] and [P. ocularis, P. aur-atus]). Additionally, a Central Hills lineage within this clade later gaverise to both a species in the Knuckles Hills (P. hoffmanni) as well as alowland species (P. hoipolloi) – members of this clade are often found inopen habitats, including anthropogenic habitats.

The common ancestor of Clade E (Fig. 2A; node 82) is estimated tohave been the result of a dispersal event from the Central Hills to theRakwana Hills. The Rakwana Hills lineage of Clade E in turn gave riseto the ancestor of two species that back-migrated to the Central Hills,and several species that now occur in the lowlands or are altitudinallywidespread. Interestingly, only clade E is represented in both Sri Lankaand India. Given that it was the lowland plains of the (now submerged)Palk Isthmus that connected Sri Lanka to India during glacial sea-levellowstands, species from clade E that are now found in India most likelyarose from a lowland ancestor. This suggests that a transition from cool-wet-adapted species to warm-dry-adapted ones occurred, facilitatingthe back migration to India during the Miocene. In fact, P. regius, thesister species to the Indian Pseudophilautus, is the only species that is

Fig. 3. Rates of diversification of Pseudophilautus. A cumulative lineage-through-time plot indicates a slow diversification rate through time without an early burst oflineages. However, a slight late burst is suggested with a gamma (γ) value of 2.91, which might be attributable to Pleistocene and Pliocene diversification. The null(Yule) model is indicated by the purple line. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of thisarticle.)

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

19

found in the dry zone of Sri Lanka. The Ps. folicola clade, which is thesister taxon to P. regius+ Indian Pseudophilautus, is also widely dis-tributed across the wet lowlands and all these species are small, leaf-litter dwellers.

Clade F (Fig. 2A; node 49) contains only three species, each in adifferent mountain range. Their MRCA is assumed, with high un-certainty, to have been a Central Hills species but all are found above800m in elevation on the three mountain regions, closely associatedwith leaf litter.

3.4. Lineage diversification in context of area and time

The gradual diversification of lineages in Sri Lanka begins duringthe Oligocene, following the initiation of the Tibetan-Himalayan oro-genic event (ca. 30 MYA), which initiated the monsoon cycle (Lichtet al., 2014; Owen and Dortch, 2014). The earliest lineages are asso-ciated with the Central Hills (Fig. 2). Early diversification of the CentralHills lineages occurs during the Mi-1 glaciation event at the Oligocene-Miocene boundary, where ice sheets formed in the Antarctic – a periodconducive to spreading of cool-wet-adapted species, such as the formsthat would have existed on the Central Hills. Lineages that can be at-tributed to Rakwana Hills arise during the early Miocene, whenephemeral Antarctic ice sheets were present (Fig. 2), which was aperiod of cooling that would have facilitated dispersal and diversifica-tion of lineages involving this region. Towards the mid Miocene, low-land lineages arise together with strengthening of the Asian monsooncycles, facilitating the genesis of lowland rainforests that would haveexpanded the available niches for the diversifying lineages. Lineagesthat cross the dry lowlands and the land bridge to India originate in thelate Miocene (Mi-4 glaciation event) with the strengthening of theAsian monsoons when for the first time Arctic ice sheets began to form,indicating a significant cooling event that would have facilitated thediversification by providing cool and wet niches across the warm anddry lowlands and the landbridge. Knuckles Hills lineages also arisebeginning in the late Miocene (Fig. 2B). The predominant pattern oflineage diversification during the Pleistocene is across adjacent moun-tain ranges (i.e., Central Hills, Rakwana Hills and Knuckles Hills), withonly one split within the lowland forms (Figs. 2, 3). This suggests thatthe increasing amplitude of climate oscillations in the late Pliocene andPleistocene alternately allowed for migration between mountain rangesduring cool, wet periods, and increased isolation among mountainranges during warm, dry periods.

4. Discussion

4.1. Timing of diversification

Adaptive radiations often are characterized by early and rapidbursts of speciation (Schluter, 2000). We find no indication of rapidinitial radiation in the Sri Lankan diversification of Pseudophilautus.Instead, we observe a relatively constant rate of diversification over theentire history of the clade, perhaps with a slight increase in recent di-versification rate. In Sri Lanka, habitat complexity provides opportu-nities for ecological and allopatric speciation. These opportunities in-clude isolated mountain ranges, a sizable tract of wet habitats, and low,dry valleys interspersing the landscape. Yet, we do not find an early-burst pattern in Pseudophilautus. The lack of an early-burst phase ofdiversification could be because climatic and habitat conditions thatprevailed at the time of initial colonization (ca. 30 MYA) differed frompresent conditions and habitats favorable for anurans were not abun-dant.

The time of orogeny of the three main mountain ranges in Sri Lankais poorly known. The few sources that refer to orogeny (Cooray, 1994;Wadia, 1945) do not provide a specific age for the Sri Lankan moun-tains. Therefore, it is not clear if mountainous regions were presentwhen the common ancestor of Pseudophilautus colonized Sri Lanka.

Interestingly, a similar diversification of Raorchestes inhabiting themountainous Western Ghats of India dates to about the same time as thediversification of Pseudophilautus (ca. 30 MYA). The branching patternof the Raorchestes phylogeny (Vijayakumar et al., 2016) also suggests arelatively constant rate of diversification and may have occurred co-incident with the uplift of the Western Ghats. If the orogeny of theWestern Ghats and mountains of Sri Lanka was roughly coincident, thediversification of both lineages may have tracked the creation of sui-table habitats as mountains were uplifted. This scenario, while spec-ulative, would explain the absence of an early-burst phase of diversi-fication and is consistent with the observation of relatively slow andconstant diversification rates in both groups. If the Western Ghats andSri Lankan mountain massifs result from the same processes of orogeny,and thus time of uplift, it may explain the coincident and gradual ra-diations of shrub frogs in both instances. Current information is in-sufficient to test this hypothesis, but the coincidence in time of the twodiversifications is striking and suggests that broad-scale, regional pro-cesses affected both Raorchestes and Pseuophilautus similarly.

4.2. Effects of climate on diversification

Climate certainly had a large effect on lineage geographic dis-tribution, dispersal and the process of speciation. The Eocene-to-Oligocene transition (EOT), underscored by the Oi-1 glaciation, likelyprovided the opportunity for colonization of Sri Lanka. The Oi-1 gla-ciation was responsible for a global cooling event that resulted in spe-cies turnover in many systems (Cornette et al., 2002; Pearson et al.,2008; Liu et al., 2009; Mayhew et al., 2012; Hull, 2015). Oi-1 was aperiod where “full scale and permanent” glaciers arose in the Antarctic(Zachos et al., 2001). Temperatures in south Asia were mild (20–25 °C)but the climate dry, as monsoon cycles were not yet fully functional(Chatterjee et al., 2013; Hodges, 2000; Kent and Muttoni, 2008;Zhisheng et al., 2011). The Oi-1 glaciation brought about a lowstand insea level (Kominz et al., 2008; Zachos et al., 2001) that would havefacilitated migration of the most recent common ancestor of Pseudo-philautus from mainland India to Sri Lanka over land bridges that wereexposed intermittently during periods of sea-level contractions(Plyusnina et al., 2016). Since this pioneering ancestor is predicted tohave been a cool-and-wet-adapted form (Fig. 2), colonization must haveoccurred when the environment on the land bridge and lowland areasof India and Sri Lanka were cool and wet. Indeed, many present-dayspecies of Raorchestes, Pseudophilautus and Mercurana inhabit cool, wetenvironments (Abraham et al., 2013; Biju and Bossuyt, 2009; Biju et al.,2010) consistent with that being the ancestral state for the clade ofPseudophilautus.

Subsequent late Miocene (8.8 MYA) back-migration ofPseudophilautus from Sri Lanka to India also coincides with a “full scaleand permanent” Antarctic Mi-4 glaciation during the middle to lateMiocene period, which would have again exposed an intervening landbridge (Wilson et al., 2009; Zachos et al., 2001). While the regionaltemperature in south Asia at the time was higher than during the lateOligocene (Chatterjee et al., 2013), the Himalayan-Tibetan orogeny waswell advanced, paving the way for two stable monsoon cycles annually(Chatterjee et al., 2013; Gehrels et al., 2003; Yin and Harrison, 2000).In contrast to the cool-wet-adapted ancestral Pseudophilautus that ori-ginally colonized Sri Lanka, the back-migrating Pseudophilautus, ac-cording to ancestral-state reconstructions, was a lowland form thatprobably was adapted to drier conditions. At present, lowland regionsof India and Sri Lanka adjacent to the shallow Palk Strait that separatesthe two land masses are warm and dry, with scrub-forest habitats inwhich a single species of Pseudophilautus occurs on either side: P. regiusin Sri Lanka and P. kani on the mainland. The ancestor of the P. re-gius+[P. kani+ P. amboli+ P. wynaadensis] clade was likely a simi-larly dry lowland-adapted species.

The period between the initial migration (Oi-1 glaciation, 31 MYA)and subsequent back-migration (Mi-4 glaciation; 8.8 MYA) coincided

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

20

with intensification of the monsoons (Briggs, 2003; Voris, 2000; Wilsonet al., 2009) and seems to have been a period favorable for diversifi-cation of Pseudophilautus on Sri Lanka. Most extant species (ca. 45) ofthe Pseudophilautus radiation are of late Miocene origin. Similar cli-matic conditions may have triggered the diversification of IndianRaorchestes, of which the majority of species similarly originated duringthe Miocene (Vijayakumar et al. 2016).

The diversification rate of Pseudophilautus in Sri Lanka shows aslight increase in the Pleistocene, with about ten recognized speciesoriginating during that period. The Pleistocene, with repeated cycles ofexpanding and contracting forests, may have facilitated rapid specia-tion in both Sri Lankan Pseudophilautus and Indian Raorchestes(Vijayakumar et al., 2016) through population expansion and dispersalduring periods of favorable climate and population isolation and di-vergence during unfavorable periods. These same influences wouldhave shifted the ranges of some Sri Lankan Pseudophilautus along alti-tudinal gradients while isolating others in lowland forest fragments(Deraniyagala, 1989, 1997, 2007).

4.3. The geography of diversification

The ancestor of Pseudophilautus and subsequent ancestral forms asrecently as the early Miocene (ca. 20 MYA) reconstruct as Central Hillsforms. Other geographic areas (Lowlands, Knuckles Hills and RakwanaHills) were colonized from the Central Hills, indicating that the CentralHills is an important center of persistence and source of migrants. TheCentral Hills is not a climatically and topographically uniform area. Dueto its size (the land area above 700m in the Central Hills is 3498 km2,which is substantially larger than both the Rakwana Hills, 182 km2, andKnuckles Hills, 565 km2) and its location at the center of the island, theCentral Hills mediates rainfall over the island during the two majormonsoons. The stronger monsoon sweeps in from the southwest,peaking in June and July; the weaker approaches from the northeast,peaking in October and November. This pattern results in a much driernorthern and eastern part of the island in comparison to southern andwestern regions (Domroes et al., 1998). Furthermore, the valleys andpeaks of the Central Hills, many of them arranged in a centrally placed,anchor-like formation, form a topographically complex area, providingopportunities for allopatric divergence within the massif.

If the orogeny in Sri Lanka did not precede the diversification event,a scenario in which Central Hills forms first and Knuckles and RakwanaHills form later can explain the existing pattern of diversification.Another, less plausible scenario is that Central, Rakwana and KnucklesHills were once part of a single mountain range but subsequently be-came isolated from one another by erosion of the Kalu and Mahawelivalleys. This scenario, however, requires that early lineages ofPseudophilautus go extinct on the Rakwana and Knuckles Hills, and itdoes not explain why the oldest lineages in the diversification are in theCentral Hills.

The geographic distribution of sister lineages in the Sri Lankan ra-diation does not show diversification clustered primarily within geo-graphic areas or among areas. Instead, cladogenic events within andbetween areas occur at about equal frequencies across the tree.Excluding the nodes for which geographic state was ambiguously re-constructed, there are 22 instances in which sister lineages arising froma node are in different areas and 26 in which sister lineages are in thesame area. This suggests that speciation occurred both within and be-tween areas and, importantly, that there were periods when climaticconditions allowed for migration between what are now climaticallyand geographically isolated mountain ranges. Species ranges likelyexpanded during favorable climatic periods and contracted during un-favorable periods, resulting in repeated instances of dispersal betweenmountain ranges followed by isolation. Thus, dynamic changes inspecies distribution and dispersal, correlated with climate fluctuations,likely were the main drivers of current species distribution and di-versity.

4.4. Patterns of historical dispersal are correlated with species naturalhistory

Species assemblages in most regions are composites of speciesdrawn from diverse clades in the phylogeny, and there is an interesting,but perhaps unsurprising, correlation among life history, current dis-tribution and the likelihood of historical dispersal among mountainranges. Overall, lineages that have lower minimum elevational rangesare more likely to be sources of dispersal. The lower minimum eleva-tional range is often associated with apparent changes in natural his-tory, such as adaptation to live among streambed and streamsideboulders, or exploitation of moist microhabitats in open areas.

Species in Clade A (Fig. 2, node 126) display little dispersal amongmountain ranges. These species are restricted to higher elevation for-ests, with only one group (the “femoralis-mooreorum” group; Fig. 2,node 120) having dispersed among mountain ranges. Species in Clade Alay clutches on leaves (as opposed to soil nesting in all other clades),and the clutches have a long pre-hatching period (37–49 days in P. fe-moralis; Bahir et al. 2005). This derived reproductive mode withinPseudophilautus may restrict members of this clade to perenially cool,wet habitats at higher elevations, thus making it less likely for lineagesto disperse among mountain ranges.

In contrast to Clade A, many sister species in Clade B (Fig. 2, node161) are distributed across adjacent mountain ranges. Several dispersalevents must be inferred to explain their distribution. This pattern sug-gests that species in Clade B have greater dispersal ability and/or lessrestrictive habitat requirements than species in Clade A. Interestingly,some species in Clade B are found in boulder habitats in proximity tostreams and streambeds (Manamendra-Arachchi and Pethiyagoda,2005; Meegaskumbura and Manamendra-Arachchi, 2005; Fig. 2, node152 and P. hallidayi). Deep crevices among large boulders providestable cool and wet microclimates that would facilitate populationpersistence in otherwise dry habitat. Boulders, streambeds and galleryforests may have provided important corridors for dispersal amongmountain ranges and into the wet lowlands for this group.

Within Clade C (Fig. 2, node 111), there are three subclades thatdisplay different ecologies and likelihood of dispersal between moun-tain ranges. Pseudophilautus microtympanum and P. steineri are range-restricted, high-altitude grassland and forest inhabitants (Manamendra-Arachchi and Pethiyagoda, 2005; Meegaskumbura and Manamendra-Arachchi, 2005). Their dispersal between mountain ranges could onlyoccur when cool, wet habitats dominated the lowlands. In contrast,species in the sordidus group are widespread and contain species cap-able of living at lower elevations in streamside boulders and vegetation(Manamendra-Arachchi and Pethiyagoda, 2005). These species arelikely to have a greater capacity to disperse along streams and galleryforests. Species in the “silus” group (P. silus, P. cf. silus and P. fulvus;Fig. 2, node 93) are found at low-to-mid elevations in closed canopyforests as well as cardamom plantations and other human modified,closed habitats (Manamendra-Arachchi and Pethiyagoda, 2005). Dis-persal for this group would be facilitated by slightly wetter conditionsthan are present currently, and dispersal could occur along galleryforests and streambeds. Finally, P. reticulatus is found in the canopy oftall, mature, lowland forests. While it is restricted to the wet southwestof Sri Lanka (Manamendra-Arachchi and Pethiyagoda, 2005) and itssister species, P. papilosus, is restricted to the Rakwana Hills(Manamendra-Arachchi and Pethiyagoda, 2005), together they aredistributed along an altitudinal gradient from the Rakwana Hills tolowland rainforests.

Species in Clade D (Fig. 2, node 44) can be considered to comprisethree subclades that are ecologically distinct and possess different dis-persal abilities. The “hoipolloi-asankai” group (P. asankai, P. pleur-otaenia, P. hoffmanni and P. hoipolloi; Fig. 2, node 41) have a widedistribution and are associated with open habitats such as forest gaps,plantations and home gardens (Manamendra-Arachchi andPethiyagoda, 2005; Meegaskumbura and Manamendra-Arachchi,

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

21

2005). They are represented on all three hill ranges and extend acrossthe lowland wet zone. Since affiliates of this clade have not dispersedinto the lowland dry zone, they are not well adapted to warm, dryconditions and may be exploiting cool, wet microhabitats within openhabitats. The subclade containing P. ocularis (narrowly distributedabove 1000m in the Rakwana Hills) and P. auratus (found from 500 to1200m; Fig. 2, node 34) constitutes species that live under forest cover.Hence, compared to the “hoipolloi-asankai” group (Fig. 2, node 41),this group shows lower vagility. The “decoris-tanu” group (P. tanu, P.singu, P. mittermeieri and P. decoris; Fig. 2, node 31) likewise is foundonly in closed forest and has a limited distribution across the RakwanaHills (P. decoris) and predominantly the lowlands (the other threespecies). The “hoipolloi-asankai” group within Clade D appears to haveevolved the ability to exploit cool, wet habitats within a warm, drymatrix, and therefore its species have greater dispersal ability.

Members of Clade E (Fig. 2, node 82) are predominantly lowlandinhabitants showing the greatest dispersal abilities and apparentadaptation to warm and dry conditions. This includes the “regius-kani”group (P. regius, P. amboli, P. wynaadensis and P. kani; Fig. 2, node 63)containing warm-dry-adapted forms, which have facilitated even theback-migration to India across the warmer and drier lowlands. Theability of these species to utilize wet microhabitats, marshes, paddyfields and riverine corridors (Manamendra-Arachchi and Pethiyagoda,2005; Meegaskumbura et al., 2007, 2009; Meegaskumbura andManamendra-Arachchi, 2005) would have facilitated their dispersal.

Species in Clade F (Fig. 2, node 49) are characterized by small bodysizes and are restricted to elevations above 800m. They are char-acterized by small distributions and show limited dispersal capabilities.These are forest cover and leaf litter-dependent forms (Manamendra-Arachchi and Pethiyagoda, 2005), which, coupled with elevationalconstraints, seem to have low dispersal abilities.

4.5. The role of terrestrial direct development in the diversification

Diversification of Pseudophilautus into a wide range of terrestrialhabitats likely was facilitated by their derived reproductive mode ofterrestrial direct development, which evolved at least 25 million yearsago (Fig. 2). Within Rhacophoridae, direct development has evolved 2or 3 times (this study supports 3 times), and in each instance the cor-responding direct-developing clade is substantially more diverse thanits indirect-developing (gel-nesting) sister clade (Meegaskumbura et al.,2015). A vast majority of species-rich, direct-developing anuran di-versifications occur in perennially wet habitats, and this is true forPseudophilautus as well. Only one species (P. regius) is found in the drierlowlands, and even there it is associated with pockets of wet habitat.

The clade of Indian Pseudophilautus that is nested within the SriLankan diversification shares most features of the island group, in-cluding direct development. Yet, it is species poor in comparison tomany other clades that diversified extensively since the late Miocene.The impressive taxonomic diversity (at both species and supra-specificlevels) and diversity of habitats occupied by anurans of the WesternGhats (Abraham et al., 2013; Biju et al., 2008, 2009, 2011, 2013, 2014;Biju and Bossuyt, 2009; Bocxlaer et al., 2010) suggest that ecologicalopportunity for Pseudophilautus back-migrating to India may have beenlimited.

In contrast, the initial, earlier migration to Sri Lanka may haveprovided abundant ecological opportunity to those early colonizers inthe form of habitat unexploited by other anurans. Yet, and surprisingly,we do not find an early burst of diversification. Rather we find a statelypace of diversification whose environmental or internal drivers remainobscure. The current exceptional diversity of Pseudophilautus within SriLanka certainly owes much to the long period of time available fordiversification (31 MYA), abundant ecological opportunity (complexhabitat and fewer competing lineages when compared to the mainland)and an associated key innovation (terrestrial direct development).

5. Conclusions

The remarkably extensive diversification of Pseudophilautus in SriLanka shows several key features. It is an ancient diversification thatbegan at the end of the Oligocene, probably facilitated by climaticevents at the Oligocene-Miocene transition. Contrary to other islandradiations, this diversification lacks an early-burst phase, suggestingperhaps that suitable habitats became available gradually over timerather than being present and available at the time of colonization. Thespecies that are adapted to lower minimum elevational ranges (i.e.,warmer and drier conditions) are more likely to be sources of dispersalamong mountain ranges when compared to the elevationally restrictedspecies. During climatically favorable periods, the valleys betweenmountain ranges were crossed by species giving rise to montane com-munities made up of members derived from disparate clades. A lineagenested within the Sri Lankan diversification back-migrated to Indiaduring the late Miocene, but this clade is depauperate, possibly due tothe prior occupation of mainland niches by other anuran lineages, in-cluding direct-developing Raorchestes. Terrestrial direct developmentmay constitute a key innovation that facilitated diversification by al-lowing species to breed away from water sources and thus to exploit awide array of terrestrial habitats. The lowland lineages in Sri Lanka arederived from montane lineages and thus the mountains of Sri Lankaacted as both species pumps and refuges throughout the 31 millionyears of evolution. Together, these features highlight the importance ofmontane regions in generating and sustaining the remarkable diversityof Sri Lankan shrub frogs through time.

Acknowledgements

We thank the following individuals and organizations for supportwith this work: Allan Larson, Thomas Stewart and two anonymous re-viewers for helpful comments that improved the manuscript; M.M.Bahir and S. Nanayakkara for fieldwork; and the Department of WildlifeConservation and Forest Department of Sri Lanka for research permits.

Grants: US National Science Foundation (DEB 0345885) to CJS andJH; National Geographic Society (7612-04) to CJS; and Society ofSystematic Biologists Graduate Student Award for Research to MM. MMwas also supported by a Harvard University Center for the Environment(HUCE) Ziff Environmental Postdoctoral Fellowship.

Ethics statement: This work was covered under Boston UniversityIACUC approval 04-006 granted to CJS.

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.ympev.2018.11.004.

References

Abraham, R.K., Pyron, R.A., Ansil, B.R., Bo Zachariah, A., Zachariah, A., 2013. Two novelgenera and one new species of treefrog (Anura: Rhacophoridae) highlight crypticdiversity in the Western Ghats of India. Zootaxa. 3640, 177–189.

AmphibiaWeb. 2018. Information on amphibian biology and conservation (web appli-cation). Available: http://amphibiaweb.org. Accessed 2018 January 20.

Bahir, M.M., Meegaskumbura, M., Manamendra-Arachchi, K., Schneider, C.J.,Pethiyagoda, R., 2005. Reproduction and terrestrial direct development in Sri Lankanshrub frogs (Amphibia: Ranidae: Rhacophorinae: Philautus). Raffles Bull. Zool. 12,339–350.

Biju, S.D., Bossuyt, F., 2009. Systematics and phylogeny of Philautus Gistel, 1848 (Anura,Rhacophoridae) in the Western Ghats of India, with descriptions of 12 new species.Zool. J. Linn. Soc. 155, 374–444.

Biju, S.D., Roelants, K., Bossuyt, F., 2008. Phylogenetic position of the montane treefrogPolypedates variabilis Jerdon, 1853 (Anura: Rhacophoridae), and description of a re-lated species. Organ. Divers. Evol. 8, 267–276.

Biju, S.D., Van Bocxlaer, I., Giri, V.B., Loader, S.P., Bossuyt, F., 2009. Two new endemicgenera and a new species of toad (Anura: Bufonidae) from the Western Ghats of India.BMC Res. Notes. 2, 241.

Biju, S.D., Bocxlaer, I., Mahony, S., Dinesh, K.P., Radhakrishnan, C., Zachariah, A., Giri,V., Bossuyt, F., 2011. A taxonomic review of the Night Frog genus Nyctibatrachus

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

22

Boulenger, 1882 in the Western Ghats, India (Anura: Nyctibatrachidae) with de-scription of twelve new species. Zootaxa. 3029, 1–96.

Biju, S.D., Shouche, Y.S., Dubois, A., Dutta, S.K., Bossuyt, F., 2010. Ground-dwellingrhacophorid frog from the highest mountain peak of the Western Ghats of India. Curr.Sci. 98 (8), 1119–1125.

Biju, S.D., Kamei, R.G., Mahony, S., Thomas, A., Garg, S., Sircar, G., Suyesh, R., 2013.Taxonomic review of the tree frog genus Rhacophorus from the Western Ghats, India(Anura: Rhacophoridae), with description of ontogenetic colour changes and re-productive behaviour. Zootaxa. 3636 (2), 257–289.

Biju, S.D., Garg, S., Gururaja, K.V., Shouche, Y., Walujkar, S.A., 2014. DNA barcodingreveals unprecedented diversity in Dancing Frogs of India (Micrixalidae, Micrixalus):a taxonomic revision with description of 14 new species. Ceylon J. Sci. Biol. Sci. 43(1), 37–123.

Blackburn, D.C., Siler, C.D., Diesmos, A.C., McGuire, J.A., Cannatella, D.C., Brown, R.M.,2013. An adaptive radiation of frogs in a Southeast Asian island archipelago.Evolution 67 (9), 2631–2646. https://doi.org/10.1111/evo.12145.

Bossuyt, F., Milinkovitch, M.C., 2001. Amphibians as indicators of early tertiary “out-of-India” dispersal of vertebrates. Science 292 (5514), 93–95.

Bossuyt, F., Meegaskumbura, M., Beenaerts, N., Gower, D.J., Pethiyagoda, R., Roelants,K., Mannaert, A., Wilkinson, M., Bahir, M.M., Manamendra-Arachchi, K., Ng, P.K.L.,Schneider, C.J., Oommen, O.V., Milinkovitch, M.C., 2004. Local endemism within theWestern Ghats-Sri Lanka biodiversity hotspot. Science 306 (5695), 479–481.

Bouckaert, R., Heled, J., Kühnert, D., Vaughan, T., Wu, C.H., Xie, D., Suchard, M.A.,Rambaut, A., Drummond, A.J., 2014. BEAST 2: a software platform for Bayesianevolutionary analysis. PLoS Comput. Biol. 10 (4), e1003537.

Briggs, J.C., 2003. The biogeographic and tectonic history of India. J. Biogeogr. 30 (3),381–388.

Cadena, C.D., Kozak, K.H., Gómez, J.P., Parra, J.L., McCain, C.M., Bowie, R.C.K.,Carnaval, A.C., Moritz, C., Rahbek, C., Roberts, T.E., Sanders, N.J., Schneider, C.J.,VanDerWal, J., Zamudio, K.R., Graham, C.H., 2011. Latitude, elevational climaticzonation, and speciation in New World vertebrates. Proc. R. Soc. Lond. B-Biol. Sci.279, 194–201.

Carnaval, A.C., Hickerson, M.J., Haddad, C.F.B., Rodrigues, M.T., Moritz, C., 2014.Stability predicts genetic diversity in the Brazilian Atlantic Forest hotspot. Science323, 785–789.

Carnaval, A.C., Hickerson, M.J., Haddad, C.F., Rodrigues, M.T., Moritz, C., 2009. Stabilitypredicts genetic diversity in the Brazilian Atlantic forest hotspot. Science 323 (5915),785–789.

Chatterjee, S., Goswami, A., Scotese, C.R., 2013. The longest voyage: tectonic, magmatic,and paleoclimatic evolution of the Indian plate during its northward flight fromGondwana to Asia. Gondwana Res. 23 (1), 238–267.

Cooray, P.G., 1994. The Precambrian of Sri Lanka: a historical review. Precambrian Res.66 (1–4), 3–18.

Cornette, J.L., Lieberman, B.S., Goldstein, R.H., 2002. Documenting a significant re-lationship between macroevolutionary origination rates and Phanerozoic pCO2 le-vels. Proc. Natl. Acad. Sci. U.S.A. 99 (12), 7832–7835.

Darriba, D., Taboada, G.L., Doallo, R., Posada, D., 2012. jModelTest 2: more models, newheuristics and parallel computing. Nat. Methods 9, 772.

Deraniyagala, S.U., 1989. The prehistory of Sri Lanka: An ecological perspective.Department of Archaeological Survey (Sri Lanka), Memoir 8, 1–366, 367–813.

Deraniyagala, S.U., 2007. The prehistory and protohistory of Sri Lanka. Central CulturalFund.

Deraniyagala, S.U., 1997. Radiocarbon chronology of Iron Age and early historicAnuradhapura, Sri Lanka: Revised age estimate. Ancient Ceylon. In: In: Taddei,Maurizio, De Marco, Giuseppe (Eds.), South Asian Archaeology. Papers from theInternational Conference of the Association of South Asian Archaeologists in WesternEurope 1997, vol. 2. Brill, Boston, pp. 759–791.

Domroes, M., Kaviani, M., Schaefer, D., 1998. An analysis of regional and intra-annualprecipitation variability over Iran using multivariate statistical methods. Theor. Appl.Climatol. 61 (3–4), 151–159.

Drummond, A.J., Rambaut, A., 2007. BEAST: Bayesian evolutionary analysis by samplingtrees. BMC Evol. Biol. 7, 214.

Drummond, A.J., Suchard, M.A., Xie, D., Rambaut, A., 2012. Bayesian phylogenetics withBEAUti and the BEAST 1.7. Mol. Biol. Evol. 29, 1969–1973.

Fjeldsa, J., Bowie, R.C.K., Rahbek, C., 2012. The role of mountain ranges in the di-versification of birds. Ann. Rev. Ecol. Evol. Syst. 43, 249–265.

Gehrels, G.E., DeCelles, P.G., Martin, A., Ojha, T.P., Pinhassi, G., Upreti, B.N., 2003.Initiation of the Himalayan Orogen as an early Paleozoic thin-skinned thrust belt.GSA Today 13, 4–9.

Graham, C.H., Ron, S.R., Santos, J.C., Schneider, C.J., Moritz, C., 2004. Integratingphylogenetics and environmental niche models to explore speciation mechanisms indendrobatid frogs. Evolution 58, 1781–1793.

Guindon, S., Gascuel, O., 2003. A simple, fast and accurate method to estimate largephylogenies by maximum-likelihood. Syst. Biol. 52, 696–704.

Harmon, L.J., Weir, J.T., Brock, C.D., Glor, R.E., Challenger, W., 2007. GEIGER: in-vestigating evolutionary radiations. Bioinformatics 24 (1), 129–131.

Hedges, S.B., Duellman, W.E., Heinicke, M.P., 2008. New World direct-developing frogs(Anura: Terrarana): molecular phylogeny, classification, biogeography, and con-servation. Zootaxa. 1737 (1), 1–182.

Hodges, K.V., 2000. Tectonics of the Himalaya and southern Tibet from two perspectives.Geol. Soc. Am. Bull. 112 (3), 324–350.

Hull, P., 2015. Life in the aftermath of mass extinctions. Curr. Biol. 25 (19), R941–R952.Kent, D.V., Muttoni, G., 2008. Equatorial convergence of India and early Cenozoic climate

trends. Proc. Natl. Acad. Sci. U.S.A. 105 (42), 16065–16070.Kominz, M.A., Browning, J.V., Miller, K.G., Sugarman, P.J., Mizintseva, S., Scotese, C.R.,

2008. Late Cretaceous to Miocene sea-level estimates from the New Jersey andDelaware coastal plain coreholes: an error analysis. Basin Res. 20 (2), 211–226.

Kozak, K.H., Wiens, J., 2006. Does niche conservatism promote speciation? A case studyin North American salamanders. Evolution 60 (12), 2604–2621.

Li, J.T., Li, Y., Klaus, S., Rao, D.Q., Hillis, D.M., Zhang, Y.P., 2013. Diversification ofrhacophorid frogs provides evidence for accelerated faunal exchange between Indiaand Eurasia during the Oligocene. Proc. Natl. Acad. Sci. U.S.A. 110 (9), 3441–3446.

Licht, A., Van Cappelle, M., Abels, H.A., Ladant, J.B., Trabucho-Alexandre, J., France-Lanord, C., Donnadieu, Y., Vandenberghe, J., Rigaudier, T., Lécuyer, C., Terry Jr, D.,2014. Asian monsoons in a late Eocene greenhouse world. Nature 513 (7519), 501.

Liu, Z., Pagani, M., Zinniker, D., DeConto, R., Huber, M., Brinkhuis, H., Shah, S.R., Leckie,R.M., Pearson, A., 2009. Global cooling during the Eocene-Oligocene climate tran-sition. Science 323 (5918), 1187–1190.

Manamendra-Arachchi, K., Pethiyagoda, R., 2005. The Sri Lankan shrub-frogs of thegenus Philautus Gistel, 1848 (Ranidae: Rhacophorinae), with description of 27 newspecies. In: Yeo, C.J., Ng, P.K.L., Pethiyagoda, R. (Eds.), Contributions to biodiversityexploration and research in Sri Lanka. Raffles B. Zool. Supplement. 12, pp. 163–303.

Manamendra-Arachchi, K., Pethiyagoda, R., 2006. Amphibians of Sri Lanka. WildlifeHeritage Trust, Colombo.

Mayhew, P.J., Bell, M.A., Benton, T.G., McGowan, A.J., 2012. Biodiversity tracks tem-perature over time. Proc. Natl. Acad. Sci. U.S.A. 109 (38), 15141–15145.

Meegaskumbura, M., Bossuyt, F., Pethiyagoda, R., Manamendra-Arachchi, K., Bahir, M.,Milinkovitch, M.C., Schneider, C.J., 2002. Sri Lanka: an amphibian hot spot. Science298, 379. https://doi.org/10.1126/science.298.5592.379.

Meegaskumbura, M., Senevirathne, G., Biju, S.D., Grag, S., Meegaskumbura, S.,Pethiyagoda, R., Hanken, J., Schneider, C.J., 2015. Patterns of reproductive modeevolution in Old World tree frogs (Anura, Rhacophoridae). Zool. Scr. 44 (5),509–522. https://doi.org/10.1111/zsc.12121.

Meegaskumbura, M., Manamendra-Arachchi, K., 2005. Description of eight new speciesof shrub frogs (Ranidae: Rhacophorinae: Philautus) from Sri Lanka. Raffles B. Zool.12, 305–338.

Meegaskumbura, M., 2007. Molecular Systematics, Evolution, and Ecology of Sri LankanShrub-frogs (Rhacophorinae: Pseudophilautus). Ph.D. dissertation. Boston University,Massachusetts, USA.

Meegaskumbura, M., Manamendra-Arachchi, K., Schneider, C.J., Pethiyagoda, R., 2007.New species amongst Sri Lanka’s extinct shrub frogs (Amphibia: Rhacophoridae:Philautus). Zootaxa. 1397, 1–15.

Meegaskumbura, M., Manamendra-Arachchi, K., Pethiyagoda, R., 2009. Two new speciesof shrub frogs (Rhacophoridae: Philautus) from the lowlands of Sri Lanka. Zootaxa.2122, 51–68.

Meegaskumbura, M., Manamendra-Arachchi, K., 2011. Two new species of shrub frogs(Rhacophoridae: Pseudophilautus) from Sri Lanka. Zootaxa. 2747, 1–18.

Meegaskumbura, M., Manamendra-Arachchi, K., Bowatte, G., Meegaskumbura, S., 2012.Rediscovery of Pseudophilautus semiruber, a diminutive shrub frog (Rhacophoridae:Pseudophilautus) from Sri Lanka. Zootaxa. 3229, 58–68.

Miller M.A., Pfeiffer W., Schwartz T., 2010. Creating the CIPRES Science Gateway forinference of large phylogenetic trees. In: Gateway Computing EnvironmentsWorkshop (GCE), IEEE, 1–8.

Moreau, R.E., 1966. The Bird Fauna of Africa and Its Islands. Academic Press, New York.Ogden, R., Thorpe, R.S., 2002. Molecular evidence for ecological speciation in tropical

habitats. Proc. Natl. Acad. Sci. U.S.A. 99, 13612–13615.Owen, L.A., Dortch, J.M., 2014. Nature and timing of Quaternary glaciation in the

Himalayan-Tibetan orogen. Quaternary Sci. Rev. 88, 14–54.Paradis, E., 2006. Analysis of Phylogenetics and Evolution with R. Springer, New York.Pearson, P.N., McMillan, I.K., Wade, B.S., Jones, T.D., Coxall, H.K., Bown, P.R., Lear,

C.H., 2008. Extinction and environmental change across the Eocene-Oligoceneboundary in Tanzania. Geology 36 (2), 179–182.

Plyusnina, E.E., Sallam, E.S., Ruban, D.A., 2016. Geological heritage of the Bahariya andFarafra oases, the central Western Desert. Egypt. J. Afr. Earth Sci. 116, 151–159.

Pybus, O., Harvey, P., 2000. Testing macro-evolutionary models using incomplete mo-lecular phylogenies. Proc. R. Soc. Lond. B-Biol. Sci. 267, 2267–2272.

Rabosky, D.L., 2006. LASER: a maximum likelihood toolkit for detecting temporal shiftsin diversification rates from molecular phylogenies. Evol. Bioinf. 2, 247–250.

Rambaut A., Suchard M.A., Xie D., Drummond, A.J., 2014. Tracer v1.6. Available: http://beast.bio.ed.ac.uk/Tracer.

Ree, R.H., Smith, S.A., 2008. Maximum likelihood inference of geographic range evolu-tion by dispersal, local extinction, and cladogenesis. Syst. Biol. 57 (1), 4–14.

Roelants, K., Jiang, J., Bossuyt, F., 2004. Endemic ranid (Amphibia: Anura) genera insouthern mountain ranges of the Indian subcontinent represent ancient frog lineages:evidence from molecular data. Mol. Phylog. Evol. 31, 730–740.

Schluter, D., 2000. The Ecology of Adaptive Radiation. Oxford University Press, Oxford.Schneider, C.J., Cunningham, M., Moritz, C., 1998. Comparative phylogeography and the

evolution of reptiles and amphibians endemic to the Wet Tropics rainforests ofAustralia. Mol. Ecol. 7, 487–498.

Schneider, C.J., Moritz, C., 1999. Rainforest refugia and evolution in Australia's WetTropics. Proc. R. Soc. Lond. B-Biol. Sci. 266, 191–196.

Schneider, C.J., Williams, S.E., 2005. Effects of Quaternary climate change on rainforestdiversity: insights from spatial analyses of species and genes in Australia's wet tropics.In: Bermingham, E., Dick, C.W., Moritz, C. (Eds.), Tropical Rainforests: Past, Present,and Future. University of Chicago Press, pp. 401–424.

Setiadi, M.I., McGuire, J.A., Brown, R.M., Zubairi, M., Iskandar, D.T., Andayani, N.,Supriatna, J., Evans, B.J., 2011. Adaptive radiation and ecological opportunity inSulawesi and Philippine fanged frog (Limnonectes) communities. Am. Nat. 178 (2),221–240.

Smith, B.T., McCormack, J.E., Cuervo, A.M., Hickerson, M.J., Aleixo, A., Cadena, C.D.,

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

23

Pérez-Emán, J., Burney, C.W., Xie, X., Harvey, M.G., Faircloth, B.C., Derryberry, E.P.,Prejean, J., Fields, S., Brumfield, R.T., 2014. The drivers of Tropical Speciation.Nature 515, 406–409.

Smith, T.B., Schneider, C.J., Holder, K., 2001. Refugial isolation versus ecological gra-dients. Genetica 112 (113), 383–439.

Stamatakis, A., 2006. RAxML-VI-HPC: maximum likelihood-based phylogenetic analyseswith thousands of taxa and mixed models. Bioinformatics 22 (21), 2688–2690.

Tamura, K., Stecher, G., Peterson, D., Filipski, A., Kumar, S., 2013. MEGA6: molecularevolutionary genetics analysis version 6.0. Mol. Biol. Evol. 30, 2725–2729.

Van Bocxlaer, I., Loader, S.P., Roelants, K., Biju, S.D., Menegon, M., Bossuyt, F., 2010.Gradual adaptation toward a range-expansion phenotype initiated the global radia-tion of toads. Science 327 (5966), 679–682.

Vijayakumar, S.P., Menezes, R.C., Jayarajan, A., Shanker, K., 2016. Glaciations, gradients,and geography: multiple drivers of diversification of bush frogs in the Western GhatsEscarpment. Proc. R. Soc. Lond. B-Biol. Sci. 283 (1836), 20161011.

Voris, H.K., 2000. Maps of Pleistocene sea levels in Southeast Asia: shorelines, riversystems and time durations. J. Biogeogr. 27 (5), 1153–1167.

Wadia, D.N., 1945. The three superposed peneplains of Ceylon. Ceylon Dept. Minerol.Prof. Pap. 1, 25–32.

Wickramasinghe, L.M., Vidanapathirana, D.R., Airyarathne, S., Rajeev, G., Chanaka, A.,Pastorini, J., Chathuranga, G., Wickramasinghe, N., 2013. Lost and found: One of theworld's most elusive amphibians, Pseudophilautus stellatus (Kelaart 1853) redis-covered. Zootaxa 3620 (1), 112–128.

Wilson, G.D., Humphrey, C.L., Colgan, D.J., Gray, K.A., Johnson, R.N., 2009. Monsoon-influenced speciation patterns in a species flock of Eophreatoicus Nicholls (Isopoda;Crustacea). Mol. Phyl. Evol. 51 (2), 349–364.

Yin, A., Harrison, T.M., 2000. Geologic evolution of the Himalayan-Tibetan orogen. Annu.Rev. Earth Planet. Sci. 28 (1), 211–280.

Zachos, J., Pagani, M., Sloan, L., Thomas, E., Billups, K., 2001. Trends, rhythms, andaberrations in global climate 65 Ma to present. Science 292 (5517), 686–693.

Zhisheng, A., Clemens, S.C., Shen, J., Qiang, X., Jin, Z., Sun, Y., Prell, W.L., Luo, J., Wang,S., Xu, H., Cai, Y., 2011. Glacial-interglacial Indian summer monsoon dynamics.Science 333 (6043), 719–723.

M. Meegaskumbura et al. Molecular Phylogenetics and Evolution 132 (2019) 14–24

24